add function 'plot_pareto_front'

src/plotting.py

@@ -483,3 +483,84 @@ def plot_fills_forager(fdf: pd.DataFrame, hlcvs_df: pd.DataFrame, start_pct=0.0,
         )
     ax.legend(legend)
     return plt
+
+
+def plot_pareto_front(df, metrics, minimize=(True, True)):
+    """
+    Plot optimization results with Pareto front highlighted.
+
+    Parameters:
+    df (pandas.DataFrame): DataFrame containing optimization results
+    metrics (tuple): Tuple of two column names to plot (metric1, metric2)
+    minimize (tuple): Tuple of booleans indicating whether each metric should be minimized (default: (True, True))
+
+    Returns:
+    matplotlib.figure.Figure: The generated plot
+    """
+    if len(metrics) != 2:
+        raise ValueError("Exactly two metrics must be provided")
+
+    metric1, metric2 = metrics
+
+    # Extract the metrics data
+    x = df[metric1].values
+    y = df[metric2].values
+
+    # Function to identify Pareto optimal points
+    def is_pareto_efficient(costs):
+        is_efficient = np.ones(costs.shape[0], dtype=bool)
+        for i, c in enumerate(costs):
+            if is_efficient[i]:
+                # Keep any point with at least one better coordinate than this one
+                if minimize[0] and minimize[1]:
+                    is_efficient[is_efficient] = np.any(costs[is_efficient] < c, axis=1)
+                elif not minimize[0] and minimize[1]:
+                    costs_comp = costs.copy()
+                    costs_comp[:, 0] = -costs_comp[:, 0]
+                    is_efficient[is_efficient] = np.any(
+                        costs_comp[is_efficient] < costs_comp[i], axis=1
+                    )
+                elif minimize[0] and not minimize[1]:
+                    costs_comp = costs.copy()
+                    costs_comp[:, 1] = -costs_comp[:, 1]
+                    is_efficient[is_efficient] = np.any(
+                        costs_comp[is_efficient] < costs_comp[i], axis=1
+                    )
+                else:  # not minimize[0] and not minimize[1]
+                    is_efficient[is_efficient] = np.any(-costs[is_efficient] < -c, axis=1)
+                is_efficient[i] = True
+        return is_efficient
+
+    # Find Pareto optimal points
+    costs = np.column_stack((x, y))
+    pareto_mask = is_pareto_efficient(costs)
+
+    # Create the plot
+    fig, ax = plt.subplots(figsize=(10, 6))
+
+    # Plot all points
+    ax.scatter(x[~pareto_mask], y[~pareto_mask], c="gray", alpha=0.5, label="Non-Pareto optimal")
+
+    # Plot Pareto optimal points
+    ax.scatter(x[pareto_mask], y[pareto_mask], c="red", label="Pareto optimal")
+
+    # Connect Pareto points with a line
+    pareto_points = costs[pareto_mask]
+    # Sort points for proper line connection
+    if minimize[0]:
+        pareto_points = pareto_points[pareto_points[:, 0].argsort()]
+    else:
+        pareto_points = pareto_points[(-pareto_points[:, 0]).argsort()]
+    ax.plot(pareto_points[:, 0], pareto_points[:, 1], "r--", alpha=0.5)
+
+    # Labels and title
+    ax.set_xlabel(metric1)
+    ax.set_ylabel(metric2)
+    ax.set_title("Optimization Results with Pareto Front")
+    ax.legend()
+
+    # Add grid
+    ax.grid(True, linestyle="--", alpha=0.6)
+
+    plt.tight_layout()
+    return fig